For access to this article, please select a purchase option:
IET members benefit from discounts to all IET publications and free access to E&T Magazine. If you are an IET member, log in to your account and the discounts will automatically be applied.
Your recommendation has been sent to your librarian.
With the promotion of AC and DC power transmission capacity, higher reliability of power equipment are required. Partial discharge detection, as a non-destructive test, plays an important part in insulation defect classification and evaluation of power equipment such as power transformers, power cables and GIS. However, Pattern recognition of partial discharge is difficult due to the high similarity of partial discharge signals under different types and different levels of defect in insulation. Consequently, we proposed a pattern recognition method based on deep learning for partial discharge. With experiments performed in laboratory, PRPD (Phase Resolved Partial Discharge) patterns under four levels of insulation defect are acquired. After pre-processing the raw data, a deep convolutional neural network (CNN) is built, trained and tuned so as to extract the deep features of PRPD patterns. With the deep learning based model, PRPD pattern under different levels of defect are classified. The results demonstrate that, compared with the traditional classification method such as SVM (Support Vector Machine), CNN based on deep learning performs better. With higher prediction accuracy, it supports reliable insulation monitoring and highly efficient planning for equipment maintenance and repair of power equipment.
Inspec keywords: partial discharges; feature extraction; power engineering computing; power apparatus; convolutional neural nets; power cable insulation; deep learning (artificial intelligence); power transformer insulation; partial discharge measurement; pattern classification; maintenance engineering
Subjects: Data handling techniques; Transformers and reactors; Power engineering computing; Dielectric breakdown and discharges; Charge measurement; Neural nets; Power cables